Date of Award
Doctor of Philosophy (PhD)
Microbiology, Immunology, and Biochemistry
Junmin Peng, Ph.D.
Suzanne J. Baker, Ph.D. Michael A. Dyer, Ph.D. David R. Nelson, Ph.D. Stanley Pounds, Ph.D. Jinghui Zhang, Ph.D.
Cancer, Liquid chromatography, Mass spectrometry, Omics integration, Proteomics, Systems biology
Mass spectrometry (MS) based-proteomics technology has been emerging as an indispensable tool for biomedical research. But the highly diverse physical and chemical properties of the protein building blocks and the dramatic human proteome complexity largely limited proteomic profiling depth. Moreover, there was a lack of high-throughput quantitative strategies that were both precise and parallel to in-depth proteomic techniques. To solve these grand challenges, a high resolution liquid chromatography (LC) system that coupled with an advanced mass spectrometer was developed to allow genome-scale human proteome identification. Using the combination of pre-MS peptide fractionation, MS2-based interference detection and post-MS computational interference correction, we enabled precise proteome quantification with isobaric labeling. We then applied these advanced proteomics tools for cancer proteome analyses on high grade gliomas (HGG) and rhabdomyosarcomas (RMS). Using systems biology approaches, we demonstrated that these newly developed proteomic analysis pipelines are able to (i) define human proteotypes that link oncogenotypes to cancer phenotypes in HGG and to (ii) identify therapeutic vulnerabilities in RMS. Development of high resolution liquid chromatography is essential for improving the sensitivity and throughput of mass spectrometry-based proteomics to genome-scale. Here we present systematic optimization of a long gradient LC-MS/MS platform to enhance protein identification from a complex mixture. The platform employed an in-house fabricated, reverse phase long column (100 µm x 150 cm, 5 µm C18 beads) coupled with Q Exactive MS. The column was capable of achieving a peak capacity of approximately 700 in a 720 min gradient of 10-45% acetonitrile. The optimal loading amount was about 6 micrograms of peptides, although the column allowed loading as many as 20 micrograms. Gas phase fractionation of peptide ions further increased the number of peptides identified by ~10%. Moreover, the combination of basic pH LC pre-fractionation with the long gradient LC-MS/MS platform enabled the identification of 96,127 peptides and 10,544 proteins at 1% protein false discovery rate in a postmortem brain sample of Alzheimer’s disease. As deep RNA sequencing of the same specimen suggested that ~16,000 genes were expressed, current analysis covered more than 60% of the expressed proteome. Isobaric labeling quantification by mass spectrometry has emerged as a powerful technology for multiplexed large-scale protein profiling, but measurement accuracy in complex mixtures is confounded by the interference from co-isolated ions, resulting in ratio compression. Here we report that the ratio compression can be essentially resolved by the combination of pre-MS peptide fractionation, MS2-based interference detection and post-MS computational interference correction. To recapitulate the complexity of biological samples, we pooled tandem mass tag (TMT) labeled E. coli peptides at 1 : 3 : 10 ratios, and added in ~20-fold more rat peptides as background, followed by the analysis of two dimensional liquid chromatography-MS/MS. Systematic investigation indicated that the quantitative interference was impacted by LC fractionation depth, MS isolation window and peptide loading amount. Exhaustive fractionation (320 x 4 h) can nearly eliminate the interference and achieve results comparable to the MS3-based method. Importantly, the interference in MS2 scans can be estimated by the intensity of contaminated y1 product ions, and we thus developed an algorithm to correct reporter ion ratios of tryptic peptides. Our data indicated that intermediate fractionation (40 x 2 h) and y1 ion-based correction allowed accurate and deep TMT protein profiling, which represents a straightforward and affordable strategy in isobaric labeling proteomics High throughput omics approaches provide an unprecedented opportunity for dissecting molecular mechanisms in cancer biology. Here we present deep profiling of whole proteome, phosphoproteome and transcriptome in two high-grade glioma mouse models driven by mutated receptor tyrosine kinase (RTK) oncogenes, platelet-derived growth factor receptor alpha (PDGFRA) and neurotrophic receptor tyrosine kinase 1 (NTRK1), analyzing 13,860 proteins (11,941 genes) and 30,431 phosphosites by mass spectrometry. Systems biology approaches identified numerous functional modules and master regulators, including 41 kinases and 26 transcription factors. Pathway activity computation and mouse survival curves indicate the NTRK1 mutation induces a higher activation of AKT targets, drives a positive feedback loop to up-regulate multiple other RTKs, and shows higher oncogenic potency than the PDGFRA mutation. Further integration of the mouse data with human HGG transcriptome data determines shared regulators of invasion and stemness. Thus, multi-omics integrative profiling is a powerful avenue to characterize oncogenic activity. There is growing emphasis on personalizing cancer therapy based on somatic mutations identified in patient’s tumors. Among pediatric solid tumors, RAS pathway mutations in rhabdomyosarcoma are the most common potentially actionable lesions. Recent success targeting CDK4/6 and MEK in RAS mutant adult cancers led our collaborator Dr. Dyer’s group to test this approach for rhabdomyosarcoma. They achieved synergistic killing of RAS mutant rhabdomyosarcoma tumor cells by combining MEK and CDK4/6 inhibitors in culture but failed to achieve efficacy in vivo using orthotopic patient derived xenografts (O-PDXs). To determine how rhabdomyosarcomas evade targeting of CDK4/6 and MEK, we collaborated to perform large-scale deep proteomic, phosphoproteomic, and epigenomic profiling of RMS tumors. Integrative analysis of these omics data detected that RMS tumor cells rapidly compensate and overcome CDK4/6 and MEK combination therapy through 6 myogenic signal transduction pathways including WNT, HH, BMP, Adenyl Cyclase, P38/MAPK and PI3K. While it is not feasible to target each of these signal transduction pathways simultaneously in RMS, we discovered that they require the HSP90 chaperone to sustain the complex developmental signal transduction milieu. We achieved specific and synergistic killing of RMS cells using sub-therapeutic concentrations of an HSP90 inhibitor (ganetespib) in combination with conventional chemotherapy used for recurrent RMS. These effects were seen in the most aggressive recurrent RMS orthotopic patient derived xenografts irrespective of RAS pathway perturbations, histologic or molecular classification. Thus, multi-omics integrative cancer profiling using our newly developed tools is powerful to identify core signaling transduction networks, tumor vulnerability (master regulators) for novel cancer therapy.
Wang, Hong (http://orcid.org/0000-0002-6215-6348), "Genome-scale Precision Proteomics Identifies Cancer Signaling Networks and Therapeutic Vulnerabilities" (2017). Theses and Dissertations (ETD). Paper 426. http://dx.doi.org/10.21007/etd.cghs.2017.0441.